Here is a link to the research paper, the demonstration video, and more examples: https://augmentedperception.github.io/deepviewvideo/. In the video, we can see that for moving objects, there is a fair amount of blurriness so I'm excited to see how far this light field video can go and how these artifacts can be minimized.

This idea reminds me of dynamic NeRF papers, such as https://www.albertpumarola.com/research/D-NeRF/index.html.